Drying stresses in cellulose nanocrystal coatings: Impact of molecular and macromolecular

The industrial implementation of cellulose nanocrystals (CNCs) in films and coatings requires thorough evaluation of the internal stresses post-consolidation, as they cause fracturing and peeling. Characterizing the impact of plasticizing additives on stress is therefore critical. Herein, we use the deflection of thin glass substrates to measure drying stresses in consolidating CNC films, and benchmark the impact of five additives (glucose, glycerol, poly(ethylene glycol) (PEG), poly(vinyl alcohol) (PVA) and bovine serum albumin). Glycerol and PEG reduced drying stresses effectively, while PEG of increased molecular weight (from 0.2 to 10 kDa), PVA, and BSA were less effective. We analyzed the temporal aspects of the process, where stress relaxation of up to 30 % was observed 2 years after coating formation. Finally, we provide a framework to evaluate the impact of CNC morphology on residual stresses. The introduced approach is expected to fast-track the optimization and implementation of coatings based on biocolloids.

While the formulations and mechanical properties of CNC films and coatings, with and without plasticizers, have been discussed in the literature (Adstedt et al., 2020;Bardet, Belgacem, & Bras, 2015;de France, Kummer, Ren, Campioni, & Nyström, 2020;Natarajan & Gilman, 2018;Wang & Walther, 2015b;Yao, Meng, Bulone, & Zhou, 2017), the development and characterization of internal stresses (Francis, Mccormick, Vaessen, & Payne, 2002) during drying remains unexplored.This is despite the expectation that the optimization of the internal stresses of a coating system, considering formulation and substrate properties, have a great impact on durability (Leterrier, 2003), gas permeability (Dumont, Tornare, Leterrier, & Månson, 2007) and strength (Tardy et al., 2019).Furthermore, films subjected to internal stresses fracture more easily than stress-free ones.This occurs as tensile residual stresses pre-load the material and thus facilitate fracture initiation and propagation upon further loading (Chiu, Garino, & Cima, 1993;Tirumkudulu & Russel, 2005).In the case of CNC coatings, residual stresses are far from negligible.For instance, 20-μm films can completely distort fabrics as thick as 100 μm or become fractured when coated on rigid substrates (Tardy et al., 2019).Therefore, there is a need to limit residual stresses, for instance, by the addition of oligomeric or polymeric additives (plasticizers).When considering additives in CNC films, it is essential to not only determine their effect on mechanical strength, but also on the internal stresses arising during drying.The effect of intercalation of polymeric plasticizers between the particles on film plasticity has been discussed in the literature related to CNC films (Adstedt et al., 2020;Bai et al., 2020;Boott, Soto, Hamad, & MacLachlan, 2021;de France et al., 2020;Thérien-Aubin, Lukach, Pitch, & Kumacheva, 2015).Moderate additive loadings have been shown not to disrupt the chiral nematic order of the films (Bardet et al., 2015;Bast et al., 2021;Lin, Raghuwanshi, Browne, Simon, & Garnier, 2021;Mu & Gray, 2014;B. Wang & Walther, 2015a;Xu et al., 2018;Yao et al., 2017).Typically, an increased plasticizer concentration improves film ductility at the cost of strength and stiffness.Furthermore, an increased moisture retention can yield enhanced plasticity.Both the functional groups in the additives, as well as their molecular weight, contribute to the overall properties of the system, although it is not yet known to what extent each of these factors contribute to stress development.Therefore, a simple and reproducible methodology to evaluate the internal stresses would be highly beneficial for future developments of biocolloidal materials.
Herein, we used the deflection of thin glass substrates to quantify the residual stresses of CNC films and to benchmark the impact of additives, such as poly(ethylene glycol) (PEG), glucose, glycerol, poly(vinyl alcohol) (PVA) and bovine serum albumin (BSA).The impact of additive loading was evaluated for PEG and glycerol, which led to a progressive stress reduction as the CNC:additive ratio changed from 10:0 to 10:4.The impact of molecular weight (Mw) was evaluated for PVA and PEG.The lowest Mw PEG (0.2 kDa) generated the highest reduction, while 10 kDa PEG and PVA were the least effective plasticizers even when compared to their higher Mw counterparts (i.e. up to 100 kDa).We used scanning electron microscopy (SEM) to correlate the internal stresses with the morphological features of fractured samples.A framework to normalize stresses to single particles is demonstrated for CNC films and the impact of aging on internal stress evolution was further investigated.The presented approach can be readily implemented across a range of biopolymeric formulations to benchmark the internal stresses that develop in nano-and microstructured films and coatings, facilitating fundamental and applied research that can lead to their commercial deployment.

Suspension preparation
The materials and the characterization methods used in this work are described in Experimental S1 and S3, respectively.Here, we briefly describe the preparation of the film-forming precursor.CNC suspensions were mixed with various water-soluble additives.In general, the additive was first fully dissolved in Milli-Q water and then mixed with a concentrated (10.37 wt%) CNC aqueous suspension using vortex stirrer (Scientific Industries Vortex genie 2), reaching CNC and additive concentrations of 5.5 wt% and 1.65 wt%, respectively.All the concentrations used in this work are expressed in weight percentage (g/g).

Drying stresses
In a typical experiment, a glass cantilever was coated with the CNC suspension and allowed to dry at 22 ± 1 • C and 20 ± 1 % RH to produce cantilever-supported CNC films (Experimental S2).The cantilever underwent deflection upon drying and, for selected samples, the full vertical deflection profile was measured over time by imaging the free end of the cantilevers (Video S1), using either an optical contact angle goniometer (KSV Instruments CAM 200) or a tensiometer (Biolin Scientific Attension Theta Flex).For other samples, the equilibrium deflection of the coated cantilevers was measured 20 ± 4 h after casting.Selected samples were also imaged (at 22 ± 1 • C and 20 % RH) after storage in ambient conditions for >2 years.Note: The relative humidity may vary between 10 % in the winter and 50 % in the summer.All measured cantilever deflections were normalized to the specific areal dry density of the CNC fraction in the film.Details are found in supporting information, Experimental S4.
For the control CNC sample, the residual stress was estimated from the cantilever deflection using Eq. 1 (Corcoran, 1969): where d is the deflection of the cantilever (m), σ is the stress in the film (N/m 2 ), E is the elastic modulus (N/m 2 ), t is the thickness (m) (Table S1), L is the length of the coated area of the cantilever (m) and ϑ is the Poisson's ratio.The subscripts s and f stand for substrate (glass cantilever) and film, respectively.Only the first term in Eq. 1 was used, as the other represent <1 % of the stress value.Experimental S6 details the measurement of stress per single CNC in a film without additives.
The assumptions associated with this estimate and related equations are presented in Experimental S6.

Drying stresses in CNC films
In the formation of particulate films and coatings, internal stresses (Francis et al., 2002) arise when solvents evaporate from particle suspensions.The dried film retains these stresses to an extent which depends on the film components and the supramolecular bonds that formed during the drying.The stress develops upon gelation during CNC assembly (Honorato-Rios et al., 2016, 2018;Mu & Gray, 2014), and result from capillary bridges between CNCs, combined with the effect of pinning on the substrate.At the macroscale, this is also seen as the elastic response of the colloidal gel.The consolidation of such a colloidal gel involves increasing capillary pressures, which promote CNC-CNC interfacial interactions (Ogawa, Nishiyama, & Mazeau, 2020;Tardy et al., 2021) that provide cohesion to the film to resist the formation of fractures.
Here, the cantilever deflection method (Stoney, 1909;Tardy et al., 2019) was employed to assess stress development, and the final residual stress.In brief, a suspension was poured onto a glass cantilever with one fixed and one free end.As the suspension dries into a solid film, the free end of the cantilever moves upwards, and the vertical position of the cantilever is monitored with time (Fig. 1a).The stress is calculated using the measured deflection with Eq. 1, with the deflection being directly proportional to the stress in the film.More information on the measurement of drying stresses is found in Discussion S1.
Three main assembly stages can be identified based on the deflection of the films.In the first, "initial drying" stage (Fig. 1a), the aqueous suspension is cast onto the substrate and water immediately starts to evaporate.With time, the interparticle distance is reduced, as the concentration increases, and the CNCs adhering to the glass substrate help to restrict the shrinkage of the contact line.This leads to a decreasing contact angle over time.The CNCs self-assemble into liquid crystalline domains (Klockars et al., 2018;Schütz et al., 2020;Wang, Hamad, & MacLachlan, 2016) and experience kinetic arrest upon gelation (Honorato-Rios et al., 2016, 2018;Mu & Gray, 2014), thereby maintaining the developed long-range chiral nematic architecture.
In the second stage, triggered by gelation, "increasing capillary forces" (Fig. 1a) acting at the menisci between CNCs in the consolidated network cause the onset of residual stresses (Ogawa et al., 2020;Tirumkudulu & Russel, 2004).Thereafter, the capillary forces drive the shrinkage of the CNC hydrogel, which can be described as being analogue to a shrinking spring whose ends are fixed to the glass substrate.Strong CNC-glass and CNC-CNC interactions restrict shrinkage, with further drying leading to increased tensile stresses in the film (Fig. 1b).As the stresses are coupled to the substrate, the cantilever experiences compressive and tensile stresses at the upper and lower sections, respectively.A bending moment is thus generated, and, because of the high length-to-width aspect ratio and low thickness of the substrate, the cantilever experiences an upward bending.
In the third, "final drying" stage (Fig. 1a), an equilibrium internal stress and the final mass of the film are reached as a function of relative humidity, as water cannot be further removed from the system.In this condition, the film is considered "dried" (Natarajan et al., 2017;O'Neill et al., 2017).The stress is retained in tensile mode due to the strong selfcohesion of the CNC network.This contrasts with other rigid particles, such as calcium carbonate (Price, Wu, McCormick, & Francis, 2015; Fig. 1.Measurement of residual stresses in drying cellulose nanocrystal films using the cantilever deflection method.a) Schematic illustration of the build-up of the residual stress in a drying film, deposited on a glass cantilever.The bending of the cantilever correlates with the stress build-up.b) Schematic illustration of the forces in the film and in the glass cantilever.The CNC film behaves as a shrinking spring, as drying-induced capillary forces cause tensile stresses in the CNCs.The cones (that attach the spring to the substrate) represent the pinned contact line of the film to the substrate, and the shortening of the blue spring represents the removal of water in the drying film.Note: The absolute stress value is an approximation given the uncertainties introduced from e.g. the estimation of the average thickness.
Fig. 1a provides the residual stresses of a drying film as a function of time.Such evaluation, in general, is method-dependent (Dumont et al., 2007;Leterrier, 2003).Hence, Figs.2-4 use the values of normalized residual stress, where the deflection is used "as is" without conversion to stress (Eq.1), after normalization to the deflection of the CNC film without additives.This enables accurate evaluation of the effect of the additives on the stresses of CNC films.All deflection data was normalized to the areal density (kg/m 2 ) of CNCs in each film, not the overall areal density (CNCs + additive) (Experimental S4).
Fig. 2b compares the deflection profiles of CNC films in the absence (gray) and presence of PEG 100k (the subscript refers to the molecular weight of the polymer) at a CNC:PEG ratio of 10:3.The overall stress was significantly reduced for CNC films containing PEG 100k , but the stress remains after reaching the "final drying" stage (Fig. 1a).The jagged data in the profile of the additive-free film shown in Fig. 2b, between 50 and 90 min, was likely caused by small fractures (Chiu et al., 1993;Tirumkudulu & Russel, 2005;Wedin et al., 2004), although cracks are not macroscopically visible in the fully dried film (Fig. 2c 2 ).In contrast, the slope for the PEG 100k -containing film (Fig. 2b) is smoother, indicating that PEG may prevent microfracture formation.PEG 100k itself did not significantly contribute to residual stresses (Fig. S4) and experienced comparatively negligible self-cohesion.Although PEG may decrease capillary forces by increasing the CNC interparticle distance and by reducing interfacial tension, the consequent reduction in residual stresses is minor considering the more than three-fold reduction of residual stress in relation to the neat CNC film.Therefore, PEG reduces the overall CNC-CNC supramolecular interactions, most noticeable after the onset of gelation.Importantly, PEG may not adsorb on CNC surfaces in the dispersion (Reid, Marway, et al., 2017) and interacts only weakly with CNCs in the dried film.To investigate whether the presence of PEG reduces the residual stress by effectively lubricating the glass surface and reducing the attractive CNC-glass interactions, glass cantilevers were coated with a monolayer of PEG 100k (Mathur & Moudgil, 1997).Deflection measurements were then carried out as for the reference CNC film without additive.As shown in Fig. S5, a PEG 100k monolayer on the glass substrate did not reduce the residual stress.
Fig. 2c 1 reveals the impact, on the internal stresses of dried films, of different additives added to the CNC suspensions prior to drying.Here, the equilibrium residual stress is reported for each additive and based on measurements performed ca.20 h after the film was dried at given relative humidity (RH = 20 ± 1 %) and temperature (T = 22 ± 1 • C).A solids weight ratio of CNCs to additive of 10:3 was used for all films (23 wt% additive loading).Adstedt et al. (2020) predicted through modelling that a 10-25 % additive concentration is optimal to improve the mechanical performance while preserving structural color.
PEG 100k , glucose and glycerol were the most effective stress reducers (>3-fold reduction).From a practical point of view, however, PEG 100k is a preferred choice given the possible leaching of small molecules (Pongjanyakul & Puttipipatkhachorn, 2007), which is further discussed in Section 3.5.Note that the central section of the film did not develop vivid structural color (Fig. 2c 2 ) which is likely a result of the light reflection occurring mainly in the UV region, as discussed in Section 3.7.Considering glucose and glycerol, the former was found more detrimental to the optical properties of the films, as can be seen in their edges, Fig. 2c 2 .This is possibly caused by a higher suspension viscosity when glucose was added (Telis, Telis-Romero, Mazzotti, & Gabas, 2007), which may cause early gelation (Mu & Gray, 2014), and reduce the time during which liquid crystal domains can reorient and merge into the well-ordered lamellar structures of the CNC films (Klockars et al., 2018).The network consolidation, i.e. gelation, at lower solids content could also lead to higher stress, as discussed by Croll (1979) and hence explain the more limited stress reduction of glucose compared to that of glycerol.
Both PVA 85-124k and BSA significantly reduced the stress, although the reduction was only about 30 %, compared to the >65 % reduction observed for glucose, PEG 100k and glycerol.The contribution of selfinteractions of PVA 14k to the stress in PVA 14k -containing CNC films was 11 % (Fig. S4).Although BSA is not generally used as a plasticizer, it provided a similar stress reduction as PVA 85-124k .This might stem from a delayed gelation concentration (Croll, 1979).The hazy reflection of the BSA-containing CNC film (Fig. 2c 2 ) might be linked to stress relief mechanisms.Ridge-like microstructures in the BSA-containing film were observed, whereas they were absent in the CNC reference film (Fig. S6).The CNC-BSA suspension used for casting had a pH of 7.08 (anionic BSA in its native, folded state) (Barbosa et al., 2010).Importantly, the standard deviation was significantly reduced in CNC films containing additives, possibly as a result of a higher thickness homogeneity and fewer defects, such as fractures and peeling.This greater homogeneity highlights the importance of using plasticizers and reducing stresses to attain more consistent films.

Effect of additive loading on residual stresses
PEG 100k and glycerol were added at varied concentrations to CNC suspensions (Fig. 3).This has also been shown in previous work for CNC films containing PEG 10k (Tardy et al., 2019).According to modelling by Adstedt et al. (2020), a 10 % additive loading (10:1 in Fig. 3) has no effects on the optical or mechanical properties, as the additives only fill the voids between the CNCs, which explains the similar residual stress measured at 10 wt% for these significantly different additives.The reduction in residual stresses at 10 wt% loading (Fig. 3) also showcases the difference between mechanical properties and residual stresses of CNC films.On the other hand, the same authors found that at above 30 % loading (10:4) the mechanical performance deteriorated and was combined with reflectance peak broadening (Bardet et al., 2015;Huang, Chen, Liang, Yang, & Shen, 2020;Lin et al., 2021;Xu et al., 2018).This places a CNC:additive ratio of 10:3 as suitable in terms of preserving mechanical and optical properties while reducing the residual stress.

Effect of polymeric additive molecular weight
The effect of molecular weight of the plasticizing additives on residual stresses was evaluated using PVA and PEG.PVA (Fig. 4a) exhibited a slight reduction in stresses when increasing the molecular weight of the additive from ca. 10 kDa to ca. 100 kDa.For PEG (Fig. 4b), using the same Mw range, the effects were similar.On relative terms, however, increasing the molecular weight of PEG reduced the stress more than in the case of PVA in the 10 to 100 kDa range.Specifically, the increase in molecular weight reduced the stresses by ca.17.6 % and 10.6 % using PEG and PVA, respectively.Very high molecular weight PEG and PVA may disrupt the chiral nematic order formation in the dried CNC films (Lin et al., 2021).Larger macromolecules lower the overlap concentration (Chong, Smith, & Zelikin, 2013), leading to larger capillaries in the CNC film and thus, by impeding the chiral nematic order, reduces drying stress.However, applications that require good chiral nematic order would not benefit from this type of order- suppressing stress reduction.SEM images (Fig. 5) reveal chiral nematic ordering in CNC films containing large polymers.Also, the above trends might fall below statistical significance given the small differences compared to the standard deviations.
In contrast to the previous trend, when decreasing the molecular weight of PEG from 10 to 0.2 kDa (Fig. 4b), the related stress reduction of 34.5 % is significant.This reduction can be associated to the higher efficiency of the smaller PEGs to disperse CNC aggregates (Cheng et al., 2015).This is possibly caused by a reduced depletion flocculation (Hubbe, Nanko, & McNeal, 2009), which would increase the extent to which smaller PEGs reduce the CNC-CNC interparticle interactions.Small molecules are also more effective in lowering the viscosity than macromolecules and thereby increase the network consolidation concentration, which further reduce the elastic response of the drying hydrogel (Croll, 1979).Therefore, small PVAs would also be expected to reduce the stress more than large ones.Kapanya et al. (2021) found that hydrogels based on poly(sodium 2-acrylamido-2-methylpropane sulfonate) and gelatin containing 5 % PEG 0.3k swelled 9 % more in water than those containing 5 % PEG 1.5k .This difference in hydration could also partly explain the lower reduction of stress for PEG 1k compared to PEG 0.2k .

Analysis of film cross-sections
To better understand the stress reduction mechanisms of different plasticizing additives, scanning electron microscopy was employed to observe structural differences between films containing different additives.Fig. 5a shows a jagged cross-section of a CNC film in the absence of additives, indicating the brittleness of the film and the presence of locally ordered large domains.The inset shows that the CNCs are organized in a chiral nematic manner.For PVA 85-124k -containing CNC films, Fig. 5b, a sharp and uniform fracture surface is obtained.At higher magnification, the upper inset reveals some plastic deformation with stretched-out polymer domains, possibly due to plastic deformation of additive-rich areas within the film.Yet, periodic chiral nematic order can also be observed (lower inset), indicating that PVA 85-124k does not adversely affect the formation of chiral nematic domains.
Fig. 5c displays a CNC film containing PEG 100k , with roughly the same molecular weight as PVA 85-124k (Fig. 5b).The fracture surfaces are different even though the values of the molecular weight are similar.The PEG 100k -containing film has a rougher fracture surface with pulled-out fibrillar features.Nevertheless, the inset indicates well-ordered chiral nematic architecture.Overall, PEG seems to be distributed within the film more effectively than PVA, at least at a high molecular weight.Furthermore, the CNC film with PEG 0.2k (Fig. 5d) has the most distinct fracture surface.Some smooth domains formed from the oligomers are visible across the cross-section and a more extensive plastic deformation is also noticeable.Furthermore, a separate top segment lacking periodic ordering, most likely consisting of PEG 0.2k , indicates that it might diffuse over time (de Cuadro et al., 2015;Pongjanyakul & Puttipipatkhachorn, 2007) given the small size of the molecule.
Finally, the residual stress state observed for the different films (Fig. 2) correlated qualitatively with the degree of plastic deformation observed in the SEM images of the fractured surfaces (Fig. 5).Furthermore, the observations from the cross-sections support the use of PEG as an effective plasticizer.PEG 0.2k , however, behaved similarly to small plasticizers, which may diffuse out of the CNC film.

Effect of aging
The cantilever-supported CNC films were stored for at least 2 years in ambient indoor conditions.Fig. 6 shows the % change in stress during this time, with most samples shown to lose approximately 20-30 % of the initial stress.Plastic deformation is not expected as a stress relief mechanism for the additive-free CNC film that consists only of rigid particles.Therefore, the stress reduction could result from submicrometric fractures.However, their macroscopic visual appearance was retained, with no apparent fracturing.
The BSA, PVAs and glucose additives caused slightly lower reductions in the stresses of the CNC films.They may have mitigated microfracture formation and propagation, which would decrease the reduction.The effect of molecular weight was minor for the PVAs, whereas it was considerable for the PEGs, as also shown in Fig. 4a.While all PEGs relaxed the stresses, the rate of the stress relaxation may depend on the molecular weight, where smaller molecular weight PEGs relaxed the stress faster.In other words, larger PEGs do not reach all stress relaxation potential immediately but over time.Interestingly, the stress increased in the CNC film containing PEG 0.2 .This may be caused by the oligomeric PEG 0.2 leaching out of the CNC film, as discussed in Section 3.5 and shown in the cross-section image of Fig. 5d.This would potentially induce a form of drying stresses, occurring over a long time compared to the evaporation of water, thus leading to increased stress in the film.The films containing glycerol also experienced an increased stress, where mechanisms similar to PEG 0.2 -containing films may have taken place.
Besides the effect of aging, other factors such as the drying rate (Beck, Bouchard, Chauve, & Berry, 2013;Natarajan et al., 2017) and thermal annealing (D'Acierno et al., 2021) are expected to affect the residual stresses.Therefore, this same methodology may serve as a valuable tool to further investigate CNC film formation and their mechanical integrity under various conditions.

Stress distribution within CNC films
As biocolloidal particles are generally polydisperse, we herein analyzed the particle size distribution and discuss the contribution of morphological aspects to residual stress development.The dimensions of the CNCs used in this work were assessed by atomic force microscopy (AFM), Fig. S7.Fig. 7a shows a representative AFM micrograph where MountainsSPIP Expert software was used to differentiate CNCs from the background.The convexity ratio (Experimental S5) was used to differentiate single CNCs (shown in yellow) from aggregates (shown in blue).The yellow population is assumed to describe the degree of individualization present in the suspension, whereas the blue population are assumed to be drying-induced aggregates.The single CNCs, however, may also be aggregates, and are therefore called "effective" CNCs (E-CNCs).For the purposes of EISA, however, the E-CNCs may function as individual CNCs, or chiral dopants, as discussed by Parton et al. (2022) and are therefore considered the elementary particles forming the CNCadditive assemblies in this work.Small CNCs may appear shorter in the images due to the low resolution imaging.Nevertheless, due to the large quantity of data (474 particles), the results are statistically sufficient.
Fig. 7b shows the length and height data for all E-CNCs.The corresponding size distributions are displayed with a log-normal fit.Fig. 7c shows the corresponding average and median values.The CNCs are relatively long, as a result of software interpretation, with a higher aspect ratio compared to the average CNCs used in previous work (Delepierre et al., 2021;Reid, Villalobos, et al., 2017).Higher aspect ratio CNCs self-assemble at lower concentrations, with lower pitch values for the chiral nematic structures in dried films, thereby reflecting blue or UV wavelengths (Honorato-Rios & Lagerwall, 2020; Parton et al., 2022;Schütz et al., 2020).This matches the weak blue or colorless reflection of the films in Figs. 2, S4.This would also explain their blue coffee-rings, which reflect redshifted or longer wavelengths compared to the central region of the film (Klockars et al., 2019;Mu & Gray, 2015).Fig. S8 shows the length and height data for all the identified particles (E-CNCs + Aggregates).The dimensions of the E-CNCs were used to estimate the residual stress for each E-CNC, depicted in Fig. 7b as a background color scale, with the corresponding values shown in Fig. 7d.The stress per E-CNC was shown to increase both as a function of length and height (and width), but the effect of the height was much greater.When increasing the height, the stress increases semi-linearly, whereas the volume of the CNC increases quadratically.Therefore, the stress per CNC unit volume is always greater at smaller widths.Hence, lower aspect ratio CNCs are subjected to a greater stress per unit volume than larger aspect ratio ones.This can be followed visually from the orange arrow in Fig. 7b, where the stress per CNC unit volume increases by 75 % while no absolute stress increase occurs.As discussed in previous work (Honorato-Rios & Lagerwall, 2020; Klockars et al., 2018), the lower aspect ratio CNCs are more likely to exist in the top sections of the films having chiral nematic order.Consequently, the stress is expected to be concentrated in the top side of the films.

Conclusions
We monitored the deflection of cantilevers after deposition of aqueous CNC suspensions to investigate the stresses that develop during drying and arise from interparticle capillary forces, as well as their related residual stress state.PEG, PVA, glucose, glycerol and BSA were used as additives, resulting in films having decreased residual stresses and brittleness.The main factors affecting the residual stress state of the films can be summarized as follows:

Fig. 2 .
Fig. 2. Comparison of the effect of additives on the residual stresses in CNC films.The CNC:additive ratio is 10:3 in b) and c).a) Illustration of the measurement of deflection from cantilevers coated with CNC films.b) Evolution of the deflection of cantilevers coated with drying CNC suspensions without any additives and with PEG 100k as an additive.c 1 ) Comparison of the effect of different additives on the residual stresses of dried CNC films, measured ca.20 h after casting the films.Deflection data in b) and c 1 ) are normalized to the average value of the "no additives" case in c 1 ).The error bars represent the standard deviations of triplicates.c 2 ) Pictures (top view) of the coated cantilevers from the corresponding measurements presented in c 1 ).The deflection data is shown on an equal CNC areal density basis.

Fig. 3 .
Fig. 3. Comparison of the residual stresses, measured through cantilever deflections, of CNC films containing various ratios of additives (PEG 100k as the left bars and glycerol as the right bars).The ratios of CNC to additive are for the dry weights.The error bars represent standard deviations of triplicates.The deflection data is shown on an equal CNC areal density basis.

Fig. 4 .
Fig. 4. Effect of molecular weight on residual stress reduction for a) PVA and for b) PEG.The CNC:additive ratio is 10:3 and the deflection data is shown on an equal CNC areal density basis.The error bars represent standard deviations of triplicates.

Fig. 5 .
Fig. 5. SEM images of cross-sections of films on glass cantilevers.The cross-section is along the width and in the middle of the film.a) CNC film without additives, where the inset shows periodic chiral nematic order.b) CNC film containing PVA 85-124k , where the upper inset shows stretched-out domains of PVA 85-124k , while the bottom inset shows the preservation of periodic order in the film, despite the presence of PVA 85-124k .c) CNC film containing PEG 100k , where the inset shows the preservation of periodic order in the film, despite the presence of PEG 100k .d) CNC film containing PEG 0.2k .The ratio of CNCs to additive (CNC:additive) is 10:3 in b), c) and d), and 10:0 in a).The dashed blue frame is used to indicate what is expected to be a PEG-rich layer.

Fig. 6 .
Fig. 6.Change (%) in residual stress between 20 h and 2 years after casting.The numbers inside the graph represent the molecular weights of PVA and PEG.The bars show average values whereas the black dots represent individual measurements and their number vary depending on the availability of stored samples.

Fig. 7 .
Fig. 7. Size distribution of CNCs.a) AFM micrograph where CNCs are classified in two populations according to MountainsSPIP Expert software and colored accordingly.The red lines indicate assumed individualized CNC, although for size distribution purposes, the E-CNCs are defined as CNCs.b) Graph showing the height and length data for each particle from the E-CNC (brown) population, where the black lines specify different aspect ratios (length/height).The orange arrow indicates no increase in absolute stress but a 75 % increase in stress per unit volume of CNC.The corresponding length (top) and height (right) distributions are included in b), while c) shows tabulated dimensions data.Stress-related values for the background color scale in b) is included in d).